Characterization of Polymers by Matrix-Assisted Laser Desorption

4562

Macromolecules 1995,28, 4562-4569

Characterization of Polymers by Matrix-Assisted Laser Desorption Ionization-Time of Flight Mass Spectrometry. End Group Determination and Molecular Weight Estimates in Poly(ethylene glycols) Giorgio Montaudo* Dipartimento di Scienze Chimiche, Universita di Catania, Viale A. Doria, 6-95125 Catania, Italy

Maurizio S. Montaudo, Concetto Puglisi, and Filippo Samperi Istituto per la Chimica e la Tecnologia dei Materiali Polimerici, Consiglio Nazionale delle Ricerche, Viale A. Doria, 6-95125 Catania, Italy Received August 29, 1994; Revised Manuscript Received January 20, 1995@

ABSTRACT: Matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDITOF MS) allows mass measurements of large molecules such as those present in synthetic and natural macromolecules. We have used a self-calibrating method for the MALDI-TOF spectra of polymers, enabling us to obtain accurate mass values. Poly(ethy1ene glycol) (PEG) samples having two different (ii) cationic PEG, kinds of end groups were used in our work: (i) anionic PEG, H-(CHz-CHz-O),-H; H-(CHz-CHz-O),-OH. The MALDI-TOF spectra recorded in reflected mode allowed unambiguous identification of the end groups present in these PEG samples. Five anionic PEG samples with very narrow molecular weight distributions (MWD) were studied. The MALDI-TOF spectra of PEGS were recorded in both linear and reflected modes. Due to the high sensitivity and the highly linear response of HIMAS (a microchannel detector equipped with a photomultiplier), measurements of MALDI-TOF spectra were used in our work to estimate the MW and MWD of the PEG samples. Our results show that the molecular weight estimates provided by MALDI-TOF measurements agree with the values obtained by conventional techniques such as GPC, osmometry, and viscometry.

Introduction Matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) is a recently introduced1,2soft ionization technique that allows desorption and ionization of very large molecules even if in complex mixtures, such as the wide molecular weight distributions (MWD) present in synthetic and natural macromolecules.3-13 The great promise of MALDI-TOF is to perform the direct identification of mass-resolved polymer chains, including intact oligomers within a MWD, measurement of MW and of MWD, and the simultaneous determination of end groups in polymer samples. This perspective has stirred much expectation among polymer scientists. The MALDI-TOF technique was originally developed for proteins,lP2and its extension to the characterization of synthetic polymers is still ~nderway.~,~-l~ The task of obtaining accurate mass values in MALDI-TOF spectra of synthetic polymeric materials is of utmost importance, since it would make possible the direct determination of the composition of the species contained in each polymer sample and also of different end groups present in the oligomer series. End group identification and determination of the actual composition (down to species present to minor amounts) of a polymer sample are crucial in polymer analysis, and their importance cannot be overemphasized. Problems may also be encountered in determining molecular weights of polymers from their MALDI-TOF spectra. The intensity of mass peaks in mixtures is proportional to the molar amount of each species, and therefore mass spectra of polymers yield the number @

Abstract published in Advance ACS Abstracts, May 1, 1995. 0024-9297/95/2228-4562$09.00/0

average MW (although other averages may be calculated using appropriate algorithms), whereas GPC yields weight averages (although other averages may be calculated also here). This has to be taken into account when results are compared from the two techniques on polydisperse samples. Furthermore, in order to give a correct MW estimate, the MALDI-TOF detector must show a constant response to ions over a wide range of mass numbers. In the following, we report a MALDI-TOF study of PEG samples having very narrow MWD and address some of the points listed above.

Experimental Section Anionic PEG Samples. PEG samples were purchased by Polymer Laboratories. They were prepared by anionic polymerization of ethylene oxide initiated with potassium alcoholate in the presence of excess alcohol. Table 1 gives the average molecular weights, obtained by different methods for the five PEG samples, as given by the sample supplier. Cationic PEG Sample. The PEG2000 sample was obtained from Fluka. It was synthesized by reacting ethylene oxide with an excess of water in the presence of acid catalyst (0.5% HzS04). MALDI Sample Preparation. 2-((4-Hydroxyphenyl)azo)benzoic acid (HABA) was used as a matrix, and tetrahydrofuran, as a solvent. In the case of PEG2000, the solvent was ethanol and the matrix was dihydroxybenzoicacid. Probe tips were loaded with 100 pmol of polymer sample and 300 nmol of matrix. MALDI-TOF Mass Spectra. A Bruker REFLEX mass spectrometer (Bruker Franzen Analytik, Bremen, Germany) was used to obtain the matrix-assisted laser desorption ionization-time of flight mass spectra. The spectrometer was equipped with a nitrogen laser (337 nm, 5 ns) and a flash ADC (time base 8 ns). The length of the drift tube is about 1.2m. The detection in the linear mode was achieved by means of a HIMAS detector. It has a venetian-blind steel dynode which

0 1995 American Chemical Society

Macromolecules, Vol. 28,No. 13, 1995 sample PEG4100 PEG7100 PEG8600 PEG12600 PE G23600

Polymer Characterization by MALDI-TOF 4563

Table 1. Molecular Weight Averages and Molecular Weight Distributions for PEG Samples Mn Mnd M,' MwlMne Muf MPg Mnh MW1 3927" 6931"

3791 6874

116OOb 2380OC

11843 21344

4100 7100 8650 12600 23600

1.05 1.03 1.03 1.04 1.06

3900 7420 8610 12790 23710

3850 7200 8600 12750 23500

3990 7190 8380 12730 21450

4040 7360 8530 12800 21900

M,IMJ 1.01 1.02 1.02 1.01 1.02

a Vapor pressure osmometry, as measured from the sample supplier. b Membrane osmometry, as measured from the sample supplier. Light scattering, as measured from the sample supplier. Intrinsic viscosity, as measured from the sample supplier. e Estimated from GPC by the sample supplier. f Most probable mass estimated from the MALDI-TOF linear spectrum. g Most probable mass estimated from the MALDI-TOF reflection spectrum. Computed from the MALDI-TOF linear spectrum by equation M , = LNGMJNL.Computed from the MALDI-TOF linear spectrum by equation M , = ZN5(ML)2/NJtfL. 1 Calculated from the values in the last two columns.

performs the ion-to-electronconversion. Ehctrons pass through a 6000 V potential difference and hit first a single-plate

microchannel detector and then a cesium iodide scintillator which surmounts the entrance window of a photomultiplier. The conversion dynode enables the detection of high-massions. The scintillator avoids the detector saturation by providing a very wide dynamic range. The HIMAS detector has low temporal resolution. This is probably due to the production of secondary ions by the conversion dynode. The detection in the reflected mode was achieved by the standard double-plate microchannel detector. The accelerating voltage was 30 kV. The laser irradiance was slightly above threshold (ca. lo6 W/cm2). Ions below m l z 350 were removed with pulsed deflection, and 100 transients were summed. The resolution at 5000 Da was 300 in the linear mode and 500 in the reflected mode.

Results and Discussion The determination of accurate masses in the MALDI spectra of synthetic polymers that may in fact be mixtures and may possess different end groups can sometimes be a difficult task.14 The TOF spectrum is initially obtained as a temporal spectrum, which reports the number of ions arriving at each time. The mass spectrum is then usually calibrated by recording the spectrum of a compound of known mass and computing the set (Q) of calibration constants. The set Q is then used to calibrate the spectra of the unknown sample. The accuracy in mass determination by this procedure is claimed4J5 to be 0.1% (i.e. 10 mass units a t 10 000). Such a low level of accuracy is related to the fact that each mass spectrum has its own set Q, which varies when passing from one spectrum to another. For end group determinations by MALDI-TOF spectrometry, the accuracy in mass determination by the external calibration procedure is not enough and one needs a n accuracy of about 0.02%. We have used a selfcalibrating method for the MALDI-TOF spectra of polydisperse polymers, which enables us to obtain an accuracy of about 0.02% for masses up to 1000020 O0O,l2 as discussed below. The time (ti) needed for a n ion of mass pcalc(i)to fly through the tube and the mass of the ion itself are related by4,5 -6

pCalc(i) = a(ti+ 10

P)2

(1)

where a and B are calibration constants, whose values must be found through a process involving tentative assignments of the peaks in the spectra and by a linear regression p r ~ c e d u r e . ~ In the case of polymeric materials, however, one can take advantage of a unique property of the latter. Considering that the mass of a homopolymer results from the sum of identical units, eq 1 can be coupled with the condition that the mass of each oligomer results from the sum of a multiple of 0,the mass of the polymer

repeat unit, and 8, the mass of the first peak considered in the process of enumerating mass spectral peaks. Then, within the same mass series, the distance between two adjacent ions is a multiple of u,and the exact mass of the ion pmulti(i), is

pmulti(i) = (i - l ) u

+0

The calibration procedure consists in varying a and /3 until pcalc equals pmulti. This is done by minimizing the sum of squares:

(3) where the summation goes from 1to n ( n is the number of peaks in the mass series). The computation of the point of minimum, P, is cumbersome. In order to circumvent this obstacle, it is useful to introduce a function SS2, similar to SS1.

It is reasonable to assume that, if SS1 has a minimum at point P, then also SS2 takes its minimum value at point P. A computer program, called CALTOF,16 was developed in order to find the minimum of SS2 in eq 4. I t accepts as input u (the mass of the polymer repeat unit) and the time of flight (ti) corresponding to peaks in the MALDI-TOF mass spectrum. I t gives as output and pmulti(i),the calibration constants the vectors pcalc(i) a and ,!I, and the plot of SS2 as a function of 8. In Figure 1 are reported four minimization curves, obtained by the calibration procedure described above. As discussed above, the procedure takes advantage of the finding that SS2 takes its minimum a t 8, the mass of the first oligomer appearing in the MALDI-TOF spectrum. Since 0 is also known (being the repeat unit of the polymer), eqs 1 and 2 can be solved and the masses can be calculated. The mass accuracy obtained was about 0.02% at resolution 500, based on the comparison between the mass values found and those expected from the polymer formula (i.e., repeat units plus end groups plus cation). The MALDI spectra of the PEG samples were calibrated by following this procedure and are discussed below. Parts a and b of Figure 2 report the MALDI-TOF mass spectra of PEG4100, measured in linear and reflected mode, respectively. The reflected spectrum (Figure 2b) shows a higher resolution (mlAm = 500) compared to that in linear mode (Figure 2a), so that two mass series can be discerned (see inset). The anionic PEG samples investigated here (Table 1) were obtained by ring-opening polymerization of ethylene oxide initiated with potassium alcoholate in the presence of excess alcohol, acting as transfer agent and MW regulator. Assuming that the reactivities of all the OH functions in the transfer process are the same,

Macromolecules, Vol. 28, No. 13, 1995

4564 Montaudo et al. 1000

2000

1

peg 4100 alfa = 844412 beta = 4

800

600-

peg 7100 alfa = 844081 beta = 22

1600

1200

.-

5

400

800

1

200 i

400

0 1550

1560

1570

1580

1890

2000

2010

2020

40 -,

2000,

peg 8650 alfa = 843992 beta = 16

peg 12600 alfa = 844380

1

1200

800 i

--

10 -~ -T11950

1460

1970

1980

' 7

9700

9720

theta (Daltons)

Figure 1. Plots of the function SS2 in eq 4 as a function of q for samples PEG4100, PEG7100, PEG8650, and PEG12600.

independent of the chain length, the number average degree of polymerization of the poly(ethy1ene oxide) formed is given by the molar ratio of monomer and alcohol involved. The reaction scheme can be written as follows:

+

CHz-CHz-O-K+

nCH2-CHz (0)

-

CH~-CH~-O-(CHZ-CH~-O),_~ -CHz-CHz-O-K+ -CHz-CHz-O-K+

CH3-CH2-O-(CH2-CH2-O)n-j

+ CH3CH20H

-

C H ~ - C H Z - O - ( C H ~ - C H ~ - O )-CHz-CHz-OH ~-~

The latter formula can be rewritten as H-(CHz-CHz0),+1-H, from which the MW of each PEG oligomer can be calculated: (44.03n 2). If the PEG ions appear as Na' or K+ adducts, then their mass shall be given by ( 4 4 . 0 3 ~ ~2 23) or ( 4 4 . 0 3 ~ ~2 39), respectively. The MALDI-TOF spectrum in Figure 2b was calibrated with our procedure. The CALTOF computer program gave as output a = 844 412, ,L? = 4.0 (eq 1). Therefore, the most intense mass series appearing in Figure 2b is associated with ions of the type: (44.03n 2 23), and the second with ions of the type (44.03n 2 39). PEG ions appear therefore as Na+ and Kf adducts in the spectra. In recent the peaks appearing in the MALDITOF spectrum of a PEG sample were measured to have substantial mass differences from those reported above, well beyond the experimental uncertainty. For instance, the 27mer was assigned mass number 1229.7 Da (instead of 1213.8 Da), and the 37mer was assigned mass number 1670.0 Da (instead of 1654.1 Da). These mass values appear to correspond to a formula (44.03n 18 231, where the PEG oligomers are terminated by hydroxy groups, and we thought therefore that they originated from a PEG sample prepared from ethylene oxide and water in the presence of a n acidic catalyst. A sample (PEG2000), obtained by the latter synthetic

+

+ +

+ +

+ + + +

+ +

method, was then analyzed (Figure 2C), and it was found to have the mass values expected for the formula ( 4 4 . 0 3 ~ ~18 23). This experiment shows that MALDI analysis can distinguish two PEG polymers terminated by different end groups. In Figure 3a,b are reported the MALDI-TOF mass spectra of PEG 7100, measured in the linear and reflected modes, respectively. As in the previous case, the reflected spectrum shows a higher resolution compared to that in linear mode, so that the two mass series corresponding to Na+ and K+ adducts of PEGS can be distinguished. The same result was obtained in the case of PEG8650 (omitted for brevity) and of PEG12600. In the latter case, shown in Figure 4a,b, the peaks in the reflected spectrum are broader than for PEG7100. As a consequence, the peaks due to K+ adducts are no more resolved from the peaks due to Na+ adducts. The peak broadening at high masses is a well-known feature of MALDI-TOF. l7 In Figure 5a,b are reported the MALDI-TOF mass spectra of PEG23600, measured in the linear and reflected modes, respectively. I t can be seen that the spectrum is not mass resolved and therefore it is impossible to distinguish the peak due to one oligomer from that of another. In fact, in this mass range, peaks are quite broad (the width of peaks in the reflected spectrum were estimated to be 50-55 Da), so that the signal due to a given oligomer is not resolved from the signals of its nearest neighbors (the spacing between neighboring peaks is 44 Da). The noise level of the reflected spectrum in Figure 5b appears to be acceptable, but this result was obtained only after digital filtering (smoothing) of the raw data. In fact, the original S/N ratio was 5 times worse. As a matter of fact, ions strike against and reflected detector, but they produce a very small signal. This is due to the characteristics of the detector, and it has been shown that the sensitivity of this detector varies with the velocity

+ +

Macromolecules, Vol. 28,No.13, 1995


Figure 2. MALDI-TOF spectrum for sample PEG4100 in the linear mode (a), for sample PEG4100 in the reflected mode (b), and for sample PEG2000 in the reflected mode (e).

Macromolecules, VoZ. 28, No. 13, 1995

4566 Montaudo et al.

I ~ ' " ~ ' ~ ' ' " ' ~ ~ " ' ' I " ' ' I ' ~ ' ' I ' ' ~ ~ ~

2000

3000

4000

5000

6000

7000

8000

9000

lo000

mlz

111 I I

b

1000

3000

4000

........................................................................ 6000 1000 8000 9000 10000 11000 12000 m l r

5000

Figure 3. MALDI-TOF spectrum for sample PEG7100 in the linear (a) and the reflected mode (b).



1600

a'i'

1

1400

1

1200

-

1000

-

800

-

600

-

a

400

6000

8000

10000

12000

14000

16000

18000

20000

22000

nlr

a.i.

'120

100

680

660

610

620

600

10000

12000

14000

Figure 4. MALDI-TOF spectrum for sample PEG12600 in the linear (a) and the reflected mode (b)

16000

m/r


4568 Montaudo et al. a.i.

li,

a 800

600

I

I

"

~

10000

a.i.

~

l

'

"

~

1

20000

15000

~

"

~

25000

1

"

'

30000

~

I

"

~

'

I

'

~

'

'

45000

40000

35000

l

~

~

~

'

I

in/2

50000

i b

650

-

640

-

630

620

610

I

5000

'

"

'

10000

1

'

'

15000

'

~

1

20000

~

'

25000

~

'

1

30000

~

~ I

~

35000

~

'

'

l

1

~

l

~

'

40000

Figure 5. MALDI-TOF spectrum for sample PEG23600 in the linear (a) and the reflected mode (b).

'

~

*

~

45000

'

~

~

7

l

s

50000

'~

'

~

mlz

l

~r

'

~

a

~


of the incoming ions.18 The difficulties associated with the detection of large masses can be circumvented by using a detector in which the conversion from ions into electrons is performed outside the microchannel plate. l9 The linear detector possesses this characteristic and is therefore well-suited for the detection of large masses. The MALDI spectra of PEG23600 in Figure 5a,b clearly slow also the appearance of a PEG dimer at m/z 47 000. Furthermore, two peaks appear in the spectra: one, less intense, a t mlz 18 500 and another at mlz 23 600. This PEG sample appears to have a bimodal distribution of M W , a fact not detected by GPC (Table 1). In the linear mode, the ratio between the intensities of the peak at 23 600 and 18 500 Da is about 5 . In the reflected mode, this ratio is instead about 2. This is due to the fact that the reflected detector response falls a t high masses. The signal-to-noise ( S N ) ratio in the linear spectrvm is good. A comparison between the mass spectra in Figures 2-5 shows that the SIN ratio of the reflected spectra worsens appreciably at high masses, wheres the S/Nratio of the linear spectra remains nearly constant with varying mass. Table 1gives the average molecular weights for the five PEG samples investigated by MALDI-TOF together with the values given by the supplier, as obtained by different methods. Due to the high sensitivity and highly linear response of the HIMAS detector (see Experimental Section), measurements of MALDI-TOF spectra in the linear mode were used in our work to estimate the MW and MWD of the PEG samples. The reflected detector in our instrument design has different characteristics from the HIMAS (see Experimental Section), and it is less useful for MW measurements since its sensitivity falls beyond about 30-40 kDa (but its resolution increases). We have recorded the MALDI-TOF spectra also in the reflected mode in order to take advantage of the higher resolution of these spectra for end group identification purposes. Nevertheless, from Table 1 it can be seen that the MALDI-TOF spectra provide wellcomparable estimates for the most probable (M,) mass values, as measured independently from the linear and reflected spectra. This fact is due to the relatively low MW values of the PEG samples investigated. Data in Table 1also show that the molecular weight estimates provided by MALDI-TOF measurements agree with the values obtained by conventional techniques. M, values derived from MALDI-TOF spectra in the linear mode agree well with the Mn values obtained from osmometry and viscosity measurements (Table 1).Also the Mpvalues estimated from GPC come close t o the Mp values observed in the MALDI-TOF spectra. It ought t o be remarked that Mn values are all lower than Mp. The reason is that the MWD is not symmetrical with respect to its maximum. In fact, MWD has a “tail” at low masses. In the case of sample PEG23600, the value Mp = 23 710 Da, estimated from the MALDI-TOF spectrum in the linear mode (Figure 5a), differs significantly from the calculated value of M , = 21 450 Da because the latter was obtained by taking into account the bimodal distribution shown by the spectrum in Figure 5a. Comparing the GPC and MALDI-TOF Mw/Mnratios in Table 1, it appears that the MALDI-TOF spectra

Polymer Characterization by MALDI-TOF 4569 yield an appreciably narrower dispersity index with respect to GPC. The disagreement may not appear serious for these samples, prepared by anionic polymerization and therefore possessing quasi-Poisson distributions. However, for samples showing Flory distributions, or even wider, MALDI-TOF spectra may yield inaccurate results.20 Summarizing our results of the characteristics of PEGS by MALDI-TOF mass spectra, we have used a calibration procedure especially useful for end group identification and have performed a comparison of MW and MWD estimates by MALDI-TOF and conventional techniques (including GPC). The PEG samples investigated here possess a very narrow MWD; we shall report later on the characterization by MALDI-TOF of polymeric samples of wider MWD.20

Acknowledgment. Partial financial support from the Italian Ministry for University and for Scientific and Technological Research (MURST) and from the National Council of Research (CNR, Rome) is gratefully acknowledged. References and Notes (1) Karas, M.; Hillenkamp, F. Anal. Chem. 1988, 60, 2299.

(2) Hillenkamp, F.; Karas, M.; Beavis, R. C.; Chait, B. T. Anal. Chem. 1991,63, 1193A. (3) Bahr, U.; Deppe, A.; Karas, M.; Hillenkamp, F.; Giessman, U. Anal. Chem. 1992, 64, 2866. (4) Cotter, R. A. Anal. Chem. 1992, 64, 1027A. (5) Bruker. User Software Manual, Reflex MALDI-TOF Bruker: Bremen, Germany, 1993. (6) Danis, P. 0.;Karr, D. E.; Mayer, F.; Holle, A.; Watson, C. H. Org. Mass Spectrom. 1992, 27, 843. Danis, P. 0.;Karr, D. E. Org. Mass Spectrom. 1993,28, 923. (7) Danis, P. 0.;Karr, D. E.; Westmoreland, D.; Piton, M. C.; Christie, D. I.; Clay, P. A.; Kable, S. H.; Gilbert, R. G. Macromolecules 1993,26, 6684. (8) Burger, H. M.; Muller, H. M.; Seebach, D.; Bornsen, K. 0.; Schar, M.; Widmer, H. M. Macromolecules 1993, 26, 4783. (9) Eggert, M.; Freitag, R. J . Polym. Sci., Part A: Polym. Chem. 1994,32, 803. (10) Visy, C.; Lukkari, J.; Kankare, J. Macromolecules 1994,27, 3322. (11) Abate, R.; Ballistreri, A.; Montaudo, G.; Garozzo, D.; Impallomeni, G.; Critchley, G.; Tanaka, K. Rapid Commun. Mass Spectrom. 1993, 7, 1033. Garozzo, D.; Spina, E.; Sturiale, L.; Montaudo, G . Rapid Commun. Mass Spectrom. 1994, 8 , 358. (12) Montaudo, G.; Montaudo, M. S.;Puglisi, C.; Samperi, F. Anal. Chem. 1994,66,4366;Rapid Commun. Mass Spectrom. 1994, 8, 981. (13) Ehring, H.; Karas, M.; Hillenkamp, F. Org. Mass Spectrom. 1992.27.472. Beavis. R. C.: Chaudharv. T.: Chait. B. T. O w . Mass’ Spectrom. 1992; 27, 156. (14) However, this is not strictlv true in the case of MALDI-FT mass spectra. See, e.g.: Castox-o, J. A.; Wilkins, C. L. Anal. Chem. 1993,65,2621. (15) . . Zaia. J.: Annan. R. S.: Biemann. K. Ranid Commun. Mass Spectrom. 1992; 6, 32.’ (16) CALTOF DaDer listinn available on reauest. (17) Hillenkamp,-F. Procegdings of the fifth ;nternational conference on biological mass spectrometry held in Kyoto 1992; Springer: Berlin, 1992; p 22. (18) Gross, M. L., Ed.; Mass spectrometv in the biological sciences: a tutorial; Kluwer: New York, 1992; p 133. (19) Spengler, B.; Kirsh, D.; Kaufmann, R.; Karas, M.; Hillenkamp, F.; Geissmann, U. Rapid Commun. Mass Spectrom. 1992, 6, 32. (20) Montaudo, G.; Montaudo, M. S.; Puglisi, C.; Samperi, F. Rapid Commun. Mass Spectrom. 1995, 9, 453. “

MA946011S

I

I

I

Characterization of Polymers by Matrix-Assisted Laser Desorption

Recommend Documents